llm-cli-gateway 1.17.4 → 1.17.5

package/dist/upstream-contracts.js

@@ -47,9 +47,7 @@ export const UPSTREAM_CLI_CONTRACTS = {
             "excludeDynamicSystemPromptSections",
             "fallbackModel",
             "jsonSchema",
-            // Phase 4 slice ζ
             "addDir",
-            // Claude 2.x session / settings / tools surface
             "noSessionPersistence",
             "settingSources",
             "settings",
@@ -101,12 +99,6 @@ export const UPSTREAM_CLI_CONTRACTS = {
                 pattern: /^[0-9]+(?:\.[0-9]+)?$/,
                 description: "Budget cap in USD",
             },
-            // NOTE: `--probe-installed` reports --max-turns as "missing from binary"
-            // because claude 2.x hides it from the `--help` body. It is nonetheless a
-            // real, accepted flag (verified: `claude --max-turns N --help` parses
-            // without an "unknown option" error, while a genuinely unknown flag errors
-            // loudly). Keep it in the contract; the probe drift here is a known
-            // help-text false-positive, not a removed flag.
             "--max-turns": { arity: "one", pattern: /^[1-9][0-9]*$/, description: "Turn cap" },
             "--effort": { arity: "one", values: EFFORT_LEVELS, description: "Reasoning effort" },
             "--exclude-dynamic-system-prompt-sections": {
@@ -127,7 +119,6 @@ export const UPSTREAM_CLI_CONTRACTS = {
             },
             "--continue": { arity: "none", description: "Continue active session" },
             "--session-id": { arity: "one", description: "Session id" },
-            // Claude 2.x session / settings / tools surface
             "--no-session-persistence": {
                 arity: "none",
                 description: "Do not persist the session to disk (ephemeral; mirrors Codex --ephemeral)",
@@ -160,16 +151,12 @@ export const UPSTREAM_CLI_CONTRACTS = {
                 expect: "fail",
             },
             {
-                // Phase 4 slice η: --fallback-model wired through prepareClaudeRequest.
                 id: "claude-fallback-model",
                 description: "Phase 4 slice η: --fallback-model accepted",
                 args: ["-p", "hello", "--fallback-model", "claude-haiku-4-5-20251001"],
                 expect: "pass",
             },
             {
-                // Phase 4 slice η: --json-schema accepts an inline JSON Schema literal
-                // (per `claude --help` example), not a path. Codex parity for
-                // structured-output validation in one slice.
                 id: "claude-json-schema",
                 description: "Phase 4 slice η: --json-schema accepts inline JSON literal",
                 args: [
@@ -183,8 +170,6 @@ export const UPSTREAM_CLI_CONTRACTS = {
                 expect: "pass",
             },
             {
-                // Phase 4 slice ζ: --add-dir wired through prepareClaudeHighImpactFlags.
-                // Repeated once per directory; each instance has arity:"one".
                 id: "claude-add-dir",
                 description: "Phase 4 slice ζ: repeated --add-dir is accepted",
                 args: ["-p", "hello", "--add-dir", "/tmp/a", "--add-dir", "/tmp/b"],
@@ -208,10 +193,6 @@ export const UPSTREAM_CLI_CONTRACTS = {
                 expect: "pass",
             },
             {
-                // Claude CLI 2.x: stream-json requires --verbose alongside --print.
-                // The gateway emits all three together; this fixture pins the combo
-                // so a future removal of --verbose breaks loudly here instead of
-                // silently at runtime against the upstream CLI.
                 id: "claude-stream-json-requires-verbose",
                 description: "Claude CLI 2.x: --output-format stream-json + --include-partial-messages + --verbose accepted together",
                 args: [
@@ -225,12 +206,6 @@ export const UPSTREAM_CLI_CONTRACTS = {
                 expect: "pass",
             },
             {
-                // Slice κ: when caller marks promptParts with cache_control, the
-                // gateway emits `-p` as a standalone flag and pipes the JSON
-                // content-blocks payload over stdin via `--input-format
-                // stream-json`. The fixture pins the exact argv combination so
-                // a future regression (re-emitting a positional prompt, dropping
-                // `--input-format`, etc.) trips loudly here.
                 id: "claude-input-format-stream-json",
                 description: "Slice κ: `-p` standalone + --input-format stream-json + --output-format stream-json + --include-partial-messages + --verbose",
                 args: [
@@ -297,15 +272,10 @@ export const UPSTREAM_CLI_CONTRACTS = {
             "images",
             "ignoreUserConfig",
             "ignoreRules",
-            // Phase 4 slice ζ
             "workingDir",
             "addDir",
         ],
         resumeOnlyFlags: ["--last", "--all"],
-        // Phase 4 slice α (v1.8.0) verified that `codex exec resume` accepts
-        // `--output-schema` and `-c` (codex-cli 0.133.0 `exec resume --help`),
-        // so they're no longer forbidden. Current resume help does not accept
-        // session-profile or working-directory policy flags.
         resumeForbiddenFlags: ["--sandbox", "-C", "--cd", "--add-dir", "--profile"],
         flags: {
             "--last": { arity: "none", description: "Resume latest session" },
@@ -368,8 +338,6 @@ export const UPSTREAM_CLI_CONTRACTS = {
                 arity: "none",
                 description: "Resume picker: show all sessions without cwd filtering",
             },
-            // The gateway emits the short form `-C`, and the advisory contract also
-            // tracks the long `--cd` alias advertised by current Codex exec help.
             "-C": {
                 arity: "one",
                 description: "Working root for the session (Phase 4 slice ζ; new sessions only)",
@@ -410,9 +378,6 @@ export const UPSTREAM_CLI_CONTRACTS = {
                 expect: "fail",
             },
             {
-                // Phase 4 slice α: --output-schema IS accepted on resume per
-                // codex-cli 0.133.0; this fixture pins the new behaviour so future
-                // contract changes can't silently regress.
                 id: "codex-resume-output-schema",
                 description: "Phase 4 slice α: --output-schema accepted on resume (codex-cli 0.133.0)",
                 args: ["exec", "resume", "--output-schema", "/tmp/schema.json", "session-id", "hello"],
@@ -521,9 +486,7 @@ export const UPSTREAM_CLI_CONTRACTS = {
             "policyFiles",
             "adminPolicyFiles",
             "attachments",
-            // Phase 4 slice γ
             "skipTrust",
-            // Auto-approve-all ergonomic alias (equivalent to approvalMode "yolo")
             "yolo",
         ],
         flags: {
@@ -623,17 +586,13 @@ export const UPSTREAM_CLI_CONTRACTS = {
             "mcpServers",
             "allowedTools",
             "disallowedTools",
-            // Phase 4 slice δ
             "maxTurns",
-            // Phase 4 slice ζ
             "workingDir",
-            // Phase 4 slice θ — Grok HIGH parity
             "sandbox",
             "rules",
             "systemPromptOverride",
             "allow",
             "deny",
-            // Grok 0.2.x context/compaction controls
             "compactionMode",
             "compactionDetail",
         ],
@@ -669,10 +628,6 @@ export const UPSTREAM_CLI_CONTRACTS = {
                 arity: "one",
                 description: "Working directory for the invocation (Phase 4 slice ζ)",
             },
-            // Phase 4 slice θ — Grok HIGH parity. `--sandbox` is freeform per
-            // `grok --help` on 0.1.210 (no `[possible values: …]` list, unlike
-            // --effort / --permission-mode / --output-format), so we register
-            // it without a `values` constraint.
             "--sandbox": {
                 arity: "one",
                 description: "Sandbox profile for filesystem + network access (Phase 4 slice θ; freeform passthrough; env: GROK_SANDBOX)",
@@ -725,7 +680,6 @@ export const UPSTREAM_CLI_CONTRACTS = {
                 arity: "optional",
                 description: "Start the session in a new git worktree, optionally named",
             },
-            // Grok 0.2.x context/compaction controls (both enum, env-backed):
             "--compaction-mode": {
                 arity: "one",
                 values: ["summary", "transcript", "segments"],
@@ -874,13 +828,10 @@ export const UPSTREAM_CLI_CONTRACTS = {
             "mcpServers",
             "allowedTools",
             "disallowedTools",
-            // Phase 4 slice γ
             "trust",
-            // Phase 4 slice δ
             "maxTurns",
             "maxPrice",
             "maxTokens",
-            // Phase 4 slice ζ
             "workingDir",
             "addDir",
         ],
@@ -896,12 +847,6 @@ export const UPSTREAM_CLI_CONTRACTS = {
                 values: ["default", "plan", "accept-edits", "auto-approve", "chat", "explore", "lean"],
                 description: "Agent/permission mode",
             },
-            // NOTE: vibe has no reasoning-effort surface. `--effort` / `--reasoning-effort`
-            // were declared speculatively (mirroring Grok) in the provider-modernisation
-            // commit but were never accepted by the CLI: vibe 2.x argparse hard-rejects them
-            // ("error: unrecognized arguments: --effort"), failing the whole request before
-            // any model call. Removed from the contract, builder, and request schema; the
-            // mistral-effort-rejected / mistral-reasoning-effort-rejected fixtures lock it in.
             "--enabled-tools": { arity: "one", description: "Enabled tool" },
             "--resume": { arity: "one", description: "Resume session" },
             "--continue": { arity: "none", description: "Continue latest session" },
@@ -916,8 +861,6 @@ export const UPSTREAM_CLI_CONTRACTS = {
             },
             "--max-price": {
                 arity: "one",
-                // Decimal-only: matches the MAX_PRICE_SCHEMA min(1e-6) lower bound
-                // that keeps String(N) in decimal form (no scientific notation).
                 pattern: /^(0|[1-9][0-9]*)(\.[0-9]+)?$/,
                 description: "Cumulative cost cap in USD (Phase 4 slice δ, programmatic mode only)",
             },
@@ -1192,24 +1135,8 @@ function validateFlagValue(cli, arg, flag, value, index, violations) {
         });
     }
 }
-/**
- * Best-effort, advisory-only extraction of long-form flags from raw --help text.
- * Returns a sorted array of unique `--foo-bar` style flags discovered in the output.
- *
- * Heuristics:
- * - Matches common option declaration lines emitted by clap, yargs, commander, custom TUIs, etc.
- * - Lowercases for stable comparison against our contract keys.
- * - Intentionally conservative: ignores obvious noise (URLs, prose in descriptions).
- *
- * This powers the bidirectional drift detector (extra flags the installed binary
- * advertises that our contract does not yet allow). It is NEVER used for argv
- * validation — only for the upstream scanner and `upstream_contracts` probe reports.
- */
 export function extractDiscoveredFlags(helpText) {
     const discovered = new Set();
-    // Long flags: --foo, --foo-bar, --foo_bar (some CLIs normalize _ to - in display).
-    // Only inspect option declaration lines so prose such as
-    // "(Claude Code: --allowedTools)" does not create false drift.
     const longRe = /--([a-z0-9][a-z0-9_-]{1,}[a-z0-9]?)/g;
     for (const line of helpText.split(/\r?\n/)) {
         const trimmed = line.trimStart();
@@ -1274,7 +1201,6 @@ export function probeInstalledCliContract(cli, timeoutMs = 5_000) {
     const discoveredFlags = extractDiscoveredFlags(helpText);
     const contractFlagSet = new Set(Object.keys(contract.flags));
     const extraFlags = discoveredFlags.filter(f => !contractFlagSet.has(f));
-    // Cheap version hint: first line that looks like a version banner
     const versionMatch = helpText.match(/^\s*(?:[A-Za-z][\w .-]+)?v?\d+\.\d+\S*/m);
     const versionHint = versionMatch ? versionMatch[0].trim().slice(0, 80) : undefined;
     const helpHash = createHash("sha256").update(helpText).digest("hex");
@@ -1303,9 +1229,6 @@ export function buildUpstreamContractReport(options = {}) {
             {
                 executable: contract.executable,
                 upstream: contract.upstream,
-                // Pure metadata pointers (changelog URLs, package name, watch
-                // categories). Enriched from the CliContract — the single source of
-                // truth — so report consumers and the scanner read the same values.
                 upstreamMetadata: contract.upstreamMetadata
                     ? {
                         sourceUrls: contract.upstreamMetadata.sourceUrls,

package/dist/validation-orchestrator.js

@@ -131,9 +131,6 @@ function plannedJudgeSynthesis(input) {
 }
 function buildProviderArgs(provider, prompt) {
     if (provider === "claude" || provider === "grok" || provider === "mistral") {
-        // Mistral Vibe mirrors Grok's `-p PROMPT` headless surface. Model selection
-        // is via VIBE_ACTIVE_MODEL env var (no --model flag); for validation runs we
-        // let the user's environment pick the active model.
         return ["-p", prompt];
     }
     if (provider === "codex")

package/dist/worktree-manager.d.ts

@@ -25,15 +25,6 @@ export declare class WorktreeCollisionError extends WorktreeError {
 }
 export declare function sanitizeWorktreeName(input: string): string;
 export declare function createWorktree(opts: CreateWorktreeOptions): Promise<WorktreeHandle>;
-/**
- * Build a SessionCleanupHook that tears down per-session worktrees. The
- * hook reads `session.metadata.worktreePath` (recorded by
- * `resolveWorktreeForRequest`) and the optional `session.metadata.worktreeName`,
- * derives `repoRoot` from the path layout (`<repoRoot>/.worktrees/<name>`),
- * and fires `removeWorktree` asynchronously. Failures are logged by
- * `removeWorktree` itself — the hook always resolves so session deletion
- * never blocks on git.
- */
 export declare function createWorktreeSessionCleanupHook(logger: Logger): (session: {
     id: string;
     metadata?: Record<string, unknown>;

package/dist/worktree-manager.js

@@ -67,7 +67,6 @@ async function execGit(repoRoot, args, logger) {
                 proc.kill("SIGKILL");
             }
             catch {
-                // ignore — process may already be gone
             }
             rejectExec(new WorktreeError(`git ${args.join(" ")} timed out after ${GIT_TIMEOUT_MS}ms`));
         }, GIT_TIMEOUT_MS);
@@ -116,8 +115,6 @@ export async function createWorktree(opts) {
     const worktreesDir = join(opts.repoRoot, ".worktrees");
     const expectedPrefix = worktreesDir + sep;
     const worktreePath = resolvePath(opts.repoRoot, ".worktrees", name);
-    // Defense in depth — sanitizeWorktreeName already blocks slashes, but
-    // double-check that the resolved path is under <repoRoot>/.worktrees/.
     if (!worktreePath.startsWith(expectedPrefix)) {
         throw new WorktreeError(`resolved worktree path escapes the expected prefix: ${worktreePath} (expected under ${expectedPrefix})`);
     }
@@ -134,8 +131,6 @@ export async function createWorktree(opts) {
         if (!registered) {
             throw new WorktreeCollisionError(worktreePath);
         }
-        // Resume reuse: the worktree already exists and is registered with git.
-        // Return a handle pointing at it without touching anything.
         logger.info?.(`reusing existing worktree at ${worktreePath}`);
         return {
             name,
@@ -145,9 +140,6 @@ export async function createWorktree(opts) {
         };
     }
     if (registered) {
-        // Path is registered but the directory is gone — let git prune it
-        // before we recreate, so `worktree add` doesn't error on the stale
-        // registration.
         const prune = await execGit(opts.repoRoot, ["worktree", "prune"], logger);
         if (prune.code !== 0) {
             logWarn(logger, `git worktree prune failed before creating ${name}: ${prune.stderr.trim()}`);
@@ -165,15 +157,6 @@ export async function createWorktree(opts) {
         createdAt: new Date().toISOString(),
     };
 }
-/**
- * Build a SessionCleanupHook that tears down per-session worktrees. The
- * hook reads `session.metadata.worktreePath` (recorded by
- * `resolveWorktreeForRequest`) and the optional `session.metadata.worktreeName`,
- * derives `repoRoot` from the path layout (`<repoRoot>/.worktrees/<name>`),
- * and fires `removeWorktree` asynchronously. Failures are logged by
- * `removeWorktree` itself — the hook always resolves so session deletion
- * never blocks on git.
- */
 export function createWorktreeSessionCleanupHook(logger) {
     return async (session) => {
         const meta = session.metadata ?? {};
@@ -181,8 +164,6 @@ export function createWorktreeSessionCleanupHook(logger) {
         if (!worktreePath)
             return;
         const worktreeName = typeof meta.worktreeName === "string" ? meta.worktreeName : undefined;
-        // Layout invariant from createWorktree: <repoRoot>/.worktrees/<name>.
-        // Strip the trailing two segments to recover repoRoot.
         const marker = `${sep}.worktrees${sep}`;
         const markerIdx = worktreePath.lastIndexOf(marker);
         if (markerIdx === -1) {
@@ -206,8 +187,6 @@ export async function removeWorktree(opts) {
         const branch = `gateway/${opts.name}`;
         const del = await execGit(opts.repoRoot, ["branch", "-D", branch], logger);
         if (del.code !== 0) {
-            // Branch may already be gone (user deleted, never existed if add
-            // half-failed, etc.). Demote to debug — this is best-effort cleanup.
             logger.debug?.(`git branch -D ${branch} returned code ${del.code}: ${del.stderr.trim()}`);
         }
     }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "llm-cli-gateway",
-  "version": "1.17.4",
+  "version": "1.17.5",
   "mcpName": "io.github.verivus-oss/llm-cli-gateway",
   "description": "MCP server providing unified access to Claude Code, Codex, Gemini, Grok, and Mistral Vibe CLIs with session management, retry logic, async job orchestration, durable job results, and cross-LLM validation.",
   "license": "MIT",